Correlated observations carry less information than they appear to. Why finance panels need clustered standard errors, how to choose the cluster, and how the readings handle one-date events and few treated firms.
Every causal claim needs a counterfactual you never observe. How propensity score matching, difference-in-differences and synthetic control build credible comparison groups, with examples from political-connection studies.
Do value stocks earn more because of what they are, or how they move? How Daniel & Titman’s characteristic-balanced sorts separate the two, and why the answer challenged factor models.
Run a regression every month, then average the slopes. How Fama–MacBeth regressions work, why each slope is secretly a portfolio return, and the pitfalls that can inflate a result.
A factor is just a long-short portfolio. How Fama & French build SMB, HML, RMW and CMA, what alpha means in a time-series regression, and what the GRS test really tells you.
Most anomalies start life as a portfolio sort. How univariate, double and triple sorts work, why NYSE breakpoints matter, and the small choices that quietly make or break a result.
Five-year “anomalies” after IPOs and equity issues look huge, until you fix the benchmark. How BHARs and calendar-time portfolios work, why they disagree, and how Bessembinder & Zhang reconcile them.
How do you measure what the market learned from a single announcement? A walkthrough of the short-window event study, from pinning down day 0 to computing CARs and fixing standard errors.
Every “abnormal return” is abnormal relative to a model. A plain-language look at Fama’s joint-hypothesis problem, and why it barely matters over days but decides everything over years.
Bayesian Vector Autoregression (BVAR) models how multiple economic variables interact over time while using Bayesian shrinkage to reduce overfitting. In this article, we explore BVAR intuitively and use leading economic indicators to forecast future ISM Manufacturing PMI.